Python 如何为pd.DataFrame实现_缺失_

我想实现

\uuuuu missing\uuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu

（与

dict

中的行为相同），而不覆盖

pandas.DataFrame

的子类。以下是我迄今为止所做的尝试：

import pandas as pd

class Bar0(pd.DataFrame):
    @property
    def _constructor(self):
        return Bar0


    def __missing__(self, key):
        print('Call __missing__ from {} with key'.format(type(self), key))
        return key


class Bar1(Bar0):
    @property
    def _constructor(self):
        return Bar1


    def __getitem__(self, key):
        print('Call __getitem__ from {} with key {}'.format(type(self), key))
        try:
            return super().__getitem__(key)
        except KeyError:
            return self.__missing__(key)


class Bar2(dict, Bar0):
    @property
    def _constructor(self):
        return Bar2


    def __init__(self, *args, **kwargs):
        Bar0.__init__(self, *args, **kwargs)
        dict.__init__(self)

        self.__getitem__ = Bar0.__getitem__


class Bar3(Bar0, dict):
    @property
    def _constructor(self):
        return Bar3


    def __init__(self, *args, **kwargs):
        Bar0.__init__(self, *args, **kwargs)
        dict.__init__(self)

        self.__getitem__ = Bar0.__getitem__

for bar in (Bar0, Bar1, Bar2, Bar3):
    foo = bar({'a': [1]})
    foo['a']
    try:
        foo['b']
    except KeyError:
        print('Raised KeyError for {}'.format(type(foo)))
    print('')

这导致：

Raised KeyError for <class '__main__.Bar0'> // Didn't call __missing__

Call __getitem__ from <class '__main__.Bar1'> with key a
Call __getitem__ from <class '__main__.Bar1'> with key b // Did call __missing__ but also __getitem__ 
Call __missing__ from <class '__main__.Bar1'> with key

Call __missing__ from <class '__main__.Bar2'> with key // Didn't even find 'a' - I guess it searches __getitem__ from dict instead of pd.DataFrame
Call __missing__ from <class '__main__.Bar2'> with key

Raised KeyError for <class '__main__.Bar3'> // Didn't call __missing__

<代码>为//引发了KeyError，但未调用\uu__ 使用键a从调用 使用键b从调用uuu getitem//Did Call uuu missing uuuuuuu，但也uuu getitem uuu 用键从中调用“缺少” 使用键//调用uuu missing uuuuuuuuuuuuu from甚至没有找到“a”-我猜它从dict而不是pd.DataFrame搜索uuuu getitem 用键从中调用“缺少” //的引发键错误未调用\u\u丢失__ 对于重要的情况：

• Python:3.5
• 熊猫：0.19.2+0.g825876c
• 操作系统：RHEL 6

---

\uuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuuu

特别是

目录中的内容。这不是一般的事情。分配给
self.\uu getitem\uu
对索引操作没有影响。还有，
\u构造函数
应该是什么？Python不会调用其构造函数
\u constructor
@user2357112，thx以获取您的评论。关于
\u构造器
（），似乎您的设计是正确的。我仍然对分配属性和定义属性的区别感到困惑。因为我假设方法本身只是一个指向函数的函数指针属性。看起来构造函数确实是一个可以识别的东西。谷歌没有找到它，因为谷歌不喜欢符号。至于分配给
self.\uuuuu getitem\uuuuu
，Python在执行特殊方法查找时跳过实例dict，因此它跳过实例dict中的
\uuuu getitem\uuuuuuu
项<代码>\uuuu缺失\uuuu

在这里是一个危险的话题。